News Aggregator

News Aggregator

Collect and cluster news coverage on any topic from multiple sources. Extracts article angles, identifies narrative trends, and surfaces what different media outlets are emphasizing.

No API keys required for basic use (uses search + fetch).

Output directory: ~/Desktop/{slug}_news_{YYYY-MM-DD}/

What It Extracts

Per Article

• Title, outlet, author, date
• Article summary (first 3-4 paragraphs)
• Key claims or angles
• Quotes from people mentioned
• Links to related coverage

Across Articles

• Common narrative angles (what everyone is saying)
• Unique angles (contrarian or exclusive takes)
• Key people/companies quoted
• Sentiment trend (positive / negative / neutral)
• Timeline of how the story evolved

Data Sources

Pipeline

Step 1: Discovery via search

TOPIC="{TOPIC}"
DATE_FILTER="after:$(date -d '7 days ago' +%Y-%m-%d)"

# Recency-weighted news
composio execute COMPOSIO_SEARCH_NEWS -d "$(jq -n --arg q "$TOPIC $DATE_FILTER" '{ query: $q }')"

# Specific angles (web search with constraints)
for ANGLE in "announcement" "analysis" "criticism OR concerns"; do
  composio execute COMPOSIO_SEARCH_WEB -d "$(jq -n --arg q "$TOPIC $ANGLE $DATE_FILTER" '{ query: $q }')"
done

# Specific outlets
for SITE in techcrunch.com theverge.com wired.com; do
  composio execute COMPOSIO_SEARCH_WEB -d "$(jq -n --arg q "site:$SITE $TOPIC" '{ query: $q }')"
done

De-duplicate citations[*].url across runs before fetching.

Step 2: Fetch article content (batch)

# Pass up to ~10 URLs per call; iterate over statuses[] for partial failures
composio execute COMPOSIO_SEARCH_FETCH_URL_CONTENT -d "$(jq -n \
  --argjson urls "$(printf '%s\n' "${ARTICLE_URLS[@]}" | jq -R . | jq -s .)" \
  '{ urls: $urls, max_characters: 12000 }')"

Step 3: Paywalled or JS-heavy outlets

When fetch returns empty or gated content:

composio execute BROWSER_TOOL_CREATE_TASK -d '{
  "task":     "Capture: headline, byline, publication date, and the first 5 body paragraphs of the article. Stop if a hard paywall blocks access — do not attempt to bypass.",
  "startUrl": "{PAYWALLED_URL}"
}'

Step 4: Cluster by Angle

Group articles by narrative:

Angle A: "Company does X, here's why it matters"
Angle B: "Critics say X has problems"
Angle C: "Industry reactions to X"
Angle D: "What X means for the future of Y"

Identify:

• Dominant narrative (what most articles say)
• Dissenting views (what 1-2 outlets push back on)
• Missing angles (what nobody is covering)

Output Format

---
topic: {TOPIC}
articles_analyzed: {N}
date_range: {start} – {end}
outlets_covered: {N}
researched_at: {ISO date}
---

## Coverage Summary

**Dominant narrative**: {what most articles say}
**Dissenting view**: {contrarian take}
**Missing angle**: {what nobody is covering}

## Articles by Angle

### Angle: {Angle Name}

#### {Article Title}
**Outlet**: {Name} | **Author**: {Name} | **Date**: {date}
**URL**: {URL}

**Summary**: {2-3 sentence summary in own words}

**Key Quote**: "{quote from article}"

**Unique claim**: {what this article says that others don't}

---

## Key People Quoted
| Person | Role | Quoted In |
|--------|------|-----------|
| {name} | {title} | {N} articles |

## Timeline
| Date | Event | Outlet |
|------|-------|--------|
| {date} | {headline} | {outlet} |

## Sentiment Distribution
- Positive: {N}% of coverage
- Neutral: {N}%
- Negative/Critical: {N}%

## Outlets Coverage Matrix
| Outlet | Articles | Angle | Sentiment |
|--------|---------|-------|-----------|
| TechCrunch | {N} | {angle} | {sentiment} |
| The Verge | {N} | {angle} | {sentiment} |

Key Use Cases

1. Media monitoring — track how your company/product is covered
2. Competitive PR — what's being written about your competitors
3. Topic research — understand a fast-moving story from multiple angles
4. Journalist mapping — who covers your beat and what angles they take
5. Crisis monitoring — track negative coverage in real time